hydrological drought forecasting using arima models (case study: karkheh basin)

the present research was planned to evaluate the skill of linear stochastic models known as arima and multiplicative seasonal autoregressive integrated moving average (sarima) model in the quantitative forecasting of the standard runoff index (sri) in karkheh basin. to this end, sri was computed in monthly and seasonal time scales in 10 hydrometric stations in 1974-75 to 2012-13 period of time and then the modeling of sri time series was done to forecast the one to six months of lead-time and up to two seasons of lead-time. the sri values related to 1974-75 to 1999-2000 were used to develop the model and the residual data (2000-2001 to 2012-13) were used in model validation. in the validation stage, the observed and the predicted values of sri were compared using correlation coefficient, error criteria and statistical tests. finally, models skills were determined in view point of forecasting of lead-time and the time scale of drought evaluation. results showed that the model accuracy in forecasting two months and one season of lead-time was high. in terms of the forecasting of sri values, the skill of sarima in monthly time scale (with a rmse and a mae of 0.61 and 0.45 respectively and a correlation coefficient average of 0.72) was better than its skill in seasonal time scale. the application of sarima in monthly time scale was therefore preferred to its application in seasonal time scale.